The Algorithm of Initial Processing of the Manuscript Image

Sayyora Nurmamatovna Iskandarova

and Shoxrux Khudayarov Turaqulov

Tashkent University of Information Technologies named after Muhammad al-Khwarizmi,

Amir Temur street, 108, 100200, Tashkent, Uzbekistan

Tashkent University of Information Technologies named after Muhammad al-Khwarizmi,

Amir Temur street, 108, 100200, Tashkent, Uzbekistan

Keywords: Recognition, Initial Processing, Pre-Processing, Relevancy Function, Normalization.

Abstract: This article presents the problems of initial image processing, the logical hardware of solving them and the

efficiency of pre-processing with the relevancy functions. The initial process has a vital role in recognition

systems. Algorithmic steps of image pre-processing based on the fuzzy sets theory are presented. Image

quality improvement and results are explained.

1 INTRODUCTION

The quality of manuscript image does not always

meet our expectations. The diversity of pho-tographic

devices and differences between their technologies

lead to image processing. Initial image processing

algorithms usually increases the recognizing

efficiency of image recognizing system by pre-

processing and noise removal. A lot of fuzzy

algorithms are used in the tasks of initial image

processing. These algorithms serve to remove excess

points from the image effec-tively, thereby increasing

the quality of different images. In recent years, there

are being con-ducted researches on making use of

fuzzy techniques in image processing in the

developed countries of the world, they are connected

with the followings:

1) The existence of advanced mathematic

devices of displaying the knowledge and processing

it;

2) They are estimated by controlling fuzziness

effectively.

Many image programs demand specialist’s expertise

in order to overcome some difficulties. The theory of

fuzzy sets and fuzzy logics are able to display human

knowledge as the fuzzy IF rules and to process. On

the other hand, while processing the image majority

difficulties are caused by the randomness and

https://orcid.org/0000-0003-3628-6146

https://orcid.org/0000-0003-0716-6145

fuzziness of the data used in the considered problems

(Mancuso et. el.,1994; Peli et.al.,1982).

There is a technique in the mathematic apparatus of

the fuzzy logics that is able to show fuzzy elements in

the color of the image more clearly, so it may be used

to increase this image’s quali-ty. The recovery of

missing parts in improving the quality of the original

image is one of the first steps in recognition problems.

Image enhancement techniques usually remove small

points and shadows, smooth regions where gray

levels do not change significantly, and cause sharp

changes in gray levels (Peng,1994).

Fuzzy logic is well-suited for building image

enhancement systems because its mathematical

framework allows knowledge of its specific

application to be incorporated in the form of rules.

This has led to the development of different image

enhancement methods based on the color of various

fuzzy logic mathematical model points. Below we

will briefly consider some of them.

2 ALGORITHM OF INITIAL

PROCESSING

Mancuso, M., Poluzzi, R. and Rizzotto, G. A.

proposed to reduce the narrowing of luminance range

dynamically with the help of fuzzy rule approach and

to pre-process it for contrast en-hancement (Mancuso

Iskandarova, S. and Turaqulov, S.

The Algorithm of Initial Processing of the Manuscript Image.

DOI: 10.5220/0011906800003612

In Proceedings of the 3rd International Symposium on Automation, Information and Computing (ISAIC 2022), pages 121-124

ISBN: 978-989-758-622-4; ISSN: 2975-9463

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

121

et. al.,1994; Peli et.al.,1982; Peng,1994). Thereby, the

necessary initial image process is carried out. The

algorithms of image quality enhancement based on

the methods developed by Peli T. and Lim are also

effec-tive (Shi,1998). In their article, Peng, S. and

Lucke propose the fuzzy filter for initial image

processing and present the filter efficiency

(Kadnichanskiy,2018). It is clear that the median

filters are capable to effectively remove the Gaussian

noise and such filters as medium filters based on the

order statistics are used to remove impulse noise. In

order to combine these two filters, we use fuzzy logic,

where-by the value l indicates the fuzzy logic

apparatus is designed (Matkovic K. et al.,2005;

Boujemaa,1992 ; Kadnichanskiy et.al. 2018).

Images taken on different devices are often displayed

on the computer with low clarity, i.e., their brightness

changes are smaller compared to their average value,

and the range is represent-ed with a large differrence

in appearance. In this case, the brightness changes not

from black to white, but from gray to light-gray. In

other words, the original range get lower than one that

is allowed. To reduce the contrast, the brightness of

the image should be increased in full range. We will

indicate image process through the image points as

following: f(x,y) and g(x,y) is initial brightness value

and the value of processed image. The image

corresponds to the screen coordinates by matching the

screen point x - row number and y - column number.

Processing the value of given point indicates the

existence of functional connection with the quality of

this image, i.e.:

𝑔(𝑥,𝑦)=𝐹(

𝑓

(𝑥,𝑦)).

The value that provides the original image quality

allows to determine the output value.

While designing and analyzing the initial image

processing systems, it would be useful to have a

description of a mathematic apparatus of the images

to be processed. There are such kind of basic

approaches, and they are the followings:

deterministic, statistics and fuzzy approaches. The

deterministic approach requires to entry a

mathematical function that describes the image and

the features of each its element will be examined. In

the statistic approach, the image is defined by the

averaged characteristics, while in fuzzy approach by

fuzzy-averaged features.

In order to implement the brightness by shape

transformation, it is of great importance to process the

digital images. With the help of it, it is possible to

correct the exposition mistakes, to divide the image’s

black or bright sides. Now, we will consider the

definition of “brightness”. In photometry, luminous

flux is identified as the scalar product of 𝑉(𝜆)

(spectral luminous efficiency function) of 𝑃



(𝜆)

(radiant power). The task of determining the contrast

is related to improving the quality of the image and

its compatibility with the display screen. If 1 byte (8

bits) is allocated for the digital representation of each

image sample, then input or output signals can take

one of 256 values. Usually, the operating range

constitutes of 0 ... 255, whereby 0 corresponds to the

black level at the time of display, and 255

corresponds to the white level. For instance, the

values of 𝑓



and 𝑓



of the original image in a

given segment a and b are correspondingly equal to

its minimal and maximal brightness.

It is convenient to examine the image in a given

segment as a fuzzy random process with the help of

logic apparatus and it provides with quality image

processing in a given segment. The random process

serves to provide a continuous image with a current

function

𝑓(𝑥,𝑦). The random process 𝑓(𝑥,𝑦) is

defined by a joint probability.

It is possible to solve this task using a point-to-point

transformation of a linear contrast in different ways,

i.e.:

The algorithm of a linear implementation of the

brightness of the image in cases where given data are

fuzzy. The membership function of 𝜇



(𝑥,𝑦)is

determined as following:

The first step - Normalization:

𝑢(𝑥,𝑦)= 𝑙

𝑓

(𝑥,𝑦) −

𝑓



𝑓

𝑚𝑖𝑛



The second step – Fuzzification:

𝜇





(𝑥,𝑦)=

𝑢(𝑥,𝑦)−𝑐



𝜎



,𝑖=1,𝑘.

The third step – Determination of fuzzification:

𝜇





(𝑥,𝑦)=

2(𝜇





(𝑥,𝑦))



,0≤𝜇





(𝑥,𝑦)≤

1−2(1−𝜇





(𝑥,𝑦))



<𝜇





(𝑥,𝑦)≤1.

In this case, 𝑐



, 𝜎



are the parameters of the

membership function.

The fourth step:

[]

max min

(, ) [ (, )]

gxy f xy

σσ

−

=⋅

−

The fifth step:

ISAIC 2022 - International Symposium on Automation, Information and Computing

122

[] []

max mi n max mi n

mi n

max mi n ma x mi n

(, ) (, )

gggg

ll ll

gggg

Mgxy g Mf xy f

ff ff

−−

=+ −

−−

The sixth step:

𝑔

(𝑥,𝑦)=

𝑓

(𝑥,𝑦) − 𝑀

𝑓

(𝑥,𝑦)

𝜎

𝑓

(𝑥,𝑦)

⋅𝜎𝑔(𝑥,𝑦)

+𝑀𝑔

𝜎𝑔

(𝑥,𝑦)

𝜎𝑓(𝑥,𝑦)

𝑓(𝑥,𝑦)+𝑀𝑔(𝑥,𝑦)

−𝑀

𝑓

(𝑥,𝑦)

𝜎𝑔

(𝑥,𝑦)

𝜎

𝑓

(𝑥,𝑦)

The seventh step:

𝑔(𝑥,𝑦)= 𝐹(

𝑓

(𝑥,𝑦))=



0,𝑔

(𝑥,𝑦)<0

𝑔

(𝑥,𝑦),0≤𝑔

(𝑥,𝑦)≤255

255,𝑔

(𝑥,𝑦)255

In the majority of cases, the images entered to the computer

are noisy, i.e. they have a smaller change in their brightness

value compared to their specified value. Therefore, the

brightness is changed not from black to white, but from gray

to gray. In other words, a real range of the brightness is

lower than its allowed value. The task of contrast

enhancement consists of extend-ing the brightness range of

the image.

This problem can be solved by replacing the linear points

with a corresponding point-by-point processing:

𝑔(𝑥,𝑦) = 𝑎𝑓(𝑥,𝑦)+𝑏

i.e. such excess α and β are obtained that serve to bring the

fuzzy values of the brightness field to certain standard

values. Here,

𝑀𝑓(𝑥,𝑦),𝜎𝑓(𝑥,𝑦)

are estimated

beforehand and the coefficients а and b are selected so that

𝑀𝑔(𝑥,𝑦),𝜎𝑔(𝑥,𝑦)

are taken for output field:

𝑔̄

(

𝑥,𝑦

)

𝑓

(

𝑥,𝑦

)

−𝑀



𝑓

(

𝑥,𝑦

)



𝜎



𝑓

(

𝑥,𝑦

)



⋅𝜎



𝑔

(

𝑥,𝑦

)



+𝑀𝑔

𝜎𝑔(𝑥,𝑦)

𝜎𝑓(𝑥,𝑦)

𝑓(𝑥,𝑦)+𝑀𝑔(𝑥,𝑦)

−𝑀

𝑓

(𝑥,𝑦)

𝜎𝑔(𝑥,𝑦)

𝜎

𝑓

(𝑥,𝑦)

i.e.:

𝑎=

𝜎𝑔(𝑥,𝑦)

𝜎

𝑓

(𝑥,𝑦)

;𝑏=𝑀𝑔(𝑥,𝑦)

−𝑀

𝑓

(𝑥,𝑦)

𝜎𝑔(𝑥,𝑦)

𝜎

𝑓

(𝑥,𝑦)

;

Here:

𝑀

𝑓

(𝑥,𝑦)=

∑

𝑓



(𝑥,𝑦) ⋅ 𝜇





(𝑥,𝑦)





∑

𝜇





(𝑥,𝑦)





𝜎

𝑓

(

𝑥,𝑦

)

=



𝑛𝑚



∑

𝑓



(𝑥,𝑦)− 𝑀𝑓



(𝑥,𝑦)



𝜇









∑

𝜇











3 CONCLUSIONS

So, while processing the images it is required to select

similar pieces of the image based on some features in

correspondence to the membership function and to

the values that are 1. The initial image processing

reduces the effect on recognizing process and

increases recognition effi-ciency.

These filtering algorithms are used in initial image

processing, because they have an im-portance in

increasing the percentage of the recognition. There

have been achieved effective results on removing

many low-quality and unnecessary points in the

images of ancient manu-scripts.

We will repeat the procedures of replacing the points

for each element of the image.

The recommended method is applied by replacing at

the points based on given segment, in this case, such

properties of the texture as similarity, roughness and

graininess are taken into account. Therefore, this

method is recommended to process the manuscript

images that have tiny details (

Figure

1).

)

The Algorithm of Initial Processing of the Manuscript Image

123

Figure 1. The result of initial image processing by using

fuzzy and statistic descriptions.

REFERENCES

Mancuso M. Poluzzi R. and Rizzotto G. A. Fuzzy filter for

dynamic range reduction and contrast enhancement,

Proc. IEEE Int. Conf. on Fuzzy Syst., IEEE Press,

Piscataway, NJ, 1994, 264-267

Peli T. and Lim J. Adaptive filtering for image

enhancement, Optical Engineering, 21, 1982, 108-112.

Peng S. and Lucke L. Fuzzy filtering for mixed noise

removal during image processing, Proc. IEEE Int. Conf.

on Fuzzy Syst., IEEE Press, Piscataway, NJ, 1994, pp.

89-93.

Shi Z. Fuzzy algorithms: With Applications to Image

Processing and Pattern Recognition. – London: Word

Scientific, 1998, pp. 225-232.

Kadnichanskiy S.A Assessment of the contrast of digital

aerial and satellite images // Geode-sy and Cartography.

- 2018. - No. 3. - pp. 46–51.

Matkovic K. et al. Global Contrast Factor – a New

Approach to Image Contrast // Computa-tional

Aesthetics, 2005. – pp. 159–168.

Boujemaa N., Stamon G., Lemoine J. and Petit E. Fuzzy

ventricular endocardiogram detec-tion with gradual

focusing decision, Proc. IEEE Int. Conf. of the

Engineering in Medicine and Biology Society, 14,

1992, 1893-1894.

Kadnichanskiy SA Assessment of the contrast of digital

aerial and satellite images // Geodesy and Cartography.

- 2018. - No. 3. - pp. 46–51.

ISAIC 2022 - International Symposium on Automation, Information and Computing

124